-
1
American Institute of Aeronautics and Astronautics
Design and Analysis of a High Speed, High Pressure Peroxide/RP-1
Turbopump
William L. Murray III1, Matthew W. Steiner
2, Joseph A. Neal
3, and Steven A. Hunt II
4
Purdue University, West Lafayette, IN, 47907
In the past, hydrogen peroxide bipropellant rocket engines were
reserved for limited-
performance applications such as propulsion for rocket-assisted
take-off (RATO) or
sounding rockets due to the use of pressure-fed propellants or
low-performance turbopump
cycles. However, research performed at Purdue University has
indicated silver screen
catalyst beds used to decompose hydrogen peroxide can operate at
much higher bed loadings
and pressures than previously believed. Thus, leveraging modern
technology to substantially
increase the performance of such engines, a single-shaft, radial
turbine, centrifugal impeller
turbopump has been designed to deliver high-pressure hydrogen
peroxide and RP-1 to a
compact and lightweight 5,000 lbf rocket engine. The turbopump
is extraordinarily small
and fast, fitting within a 150 in3 envelope, spinning at 90,000
RPM, and delivering
propellants at pressures close to 6,000 psig. Benchmarked
against the Bristol-Siddeley 605
RATO engine, the developed engine delivers a 50% greater
thrust-to-weight ratio,
confirming that modern technology and design tools can be used
to produce rocket engines
that greatly outperform heritage systems of similar
configuration.
Nomenclature
= Weighted Parameter = Rothalpy, Pressure = Gravitational
Acceleration = Fluid Velocity in Absolute Reference Frame = Fluid
Velocity in Relative Reference Frame = Fluid Density = Wheel Speed
N = Wheel speed, RPM
= Mass Flow Ratio between Primary and Secondary Flows = Net
Positive Suction Head NSS = Suction Specific Speed
Q = Volumetric Flow Rate
k = Nondimensional Cavitation Parameter
V = Fluid Axial Velocity
Subscripts
= Absolute Reference Frame = Relative Reference Frame =
Available = Required = Stagnation = Relative Reference Frame =
Tangential Component = Inlet = Outlet
1 Student, School of Aeronautics and Astronautics, Purdue
University, Indiana, AIAA Member
2 Student, School of Aeronautics and Astronautics, Purdue
University, Indiana, AIAA Member
3 Student, School of Aeronautics and Astronautics, Purdue
University, Indiana, AIAA Member
4 Student, School of Aeronautics and Astronautics, Purdue
University, Indiana, AIAA Member
-
2
American Institute of Aeronautics and Astronautics
I. Introduction
ydrogen peroxide (H2O2) bipropellant rocket engines have been
used since the 1950s in launch vehicles,
sounding rockets, and as RATO (Rocket-Assisted Take Off) units
for aircraft. However, most of these engines
have limited performance due to the use of pressure-fed
propellants, open-cycle turbopumps, or low-pressure
closed-cycle turbopumps combined with low chamber pressures. To
mitigate these performance shortcomings, a
viable single-shaft, radial turbine, centrifugal impeller
turbopump system has been developed for use with 90%
H2O2 and RP-1 propellants. The following report describes the
design study for the turbopump, which feeds a
closed-cycle 5,000 lbf engine that utilizes two high-pressure
H2O2 catalyst beds as gas generators, and generates
thrust through a H2O2 transpiration-cooled combustion chamber
with an ablative nozzle extension. While many
elements of the engine design require additional development
prior to hardware testing, this study has indicated
through the use of modern day manufacturing technologies and
modeling capabilities, a very high pressure
peroxide/RP-1 pump-fed engine can be constructed that greatly
outperforms previously-built engines of a similar
configuration.
This turbopump is extraordinarily small compared to conventional
designs, as many rockets in this thrust class
can use simpler and less expensive pressure-fed systems, rather
than a pump, to deliver propellants to the
combustion chamber.2 Many assumptions and equations used for
pump design rely heavily on experimental data, all
of which is based on other pumps and sometimes scaled to pumps
of similar design. If the turbopump is shown to be
completely outside the range of other similar pump designs, it
must be taken into consideration that some of the
experimental data used in this pump design may have a
significant error margin.
II. Background
In a conventional catalyst bed for rocket applications, H2O2
flows over dense silver-plated screens and
decomposes into high temperature (approx. 1400F) gaseous oxygen
and steam, which can be used as a
monopropellant to turn a turbine, or can be routed into a
combustion chamber to oxidize and combust a fuel. This is
the case in the Bristol-Siddeley 605 (BS-605) RATO engine, which
served as the inspiration for the design outlined
in this paper because of design similarities and because one was
available at the Maurice J. Zucrow Labs. The BS-
605, using an open gas generator cycle, decomposes H2O2 through
a silver screen catalyst bed to generate gaseous
oxygen and steam to spin a turbine, which drives pumps to
deliver liquid H2O2 and kerosene to the combustion
chamber. The turbine gas, generated from a relatively low volume
of tap-off H2O2, is expelled from the vehicle
through a small nozzle, and does not enter the combustion
chamber.
Substantial research has been performed by Purdue University
students in the past decade to develop advanced
catalyst beds for the decomposition of rocket grade hydrogen
peroxide (RGHP), defined here as H2O2 of greater
than 85% concentration in water. In 2007, a group of Purdue
students developed a silver screen catalyst bed for use
with high-pressure rocket combustion chambers operating at
chamber pressures up to 4000 psia. It was found that
higher pressures correspond to a lower pressure drop through the
catalyst bed. Additionally, high pressures were
found to yield shorter start transients, and allow for a more
compact design of the combustion chamber.1
The findings of this research prompted Prof. Stephen Heister,
Director of Purdues Maurice J. Zucrow Propulsion Laboratories, to
offer a complementary graduate level course in the design of a
turbopump to feed high-
pressure H2O2 through silver screen catalyst beds and into a
combustion chamber to be reacted with rocket-grade
kerosene. The class, categorized as a 590-level
Design/Build/Test course, would have two overall objectives: to
increase student and faculty understanding of the turbopump
design process, and to work toward developing a
research-focused turbopump test capability at the Maurice J.
Zucrow Labs. The educational mission was the primary
focus of the course; the desire to build a test capability grew
out of the overall success and interest that was shown in
the project.
III. Objectives and Requirements
At a high level, the turbopump design objective was to develop a
high-pressure, lightweight, compact, and cost-
effective turbopump to integrate with an engine propulsion
system as a launcher for a nominal micro-satellite launch
mission. At a lower level, various overall engine requirements
and goals, which affected the turbopump design, were
laid out prior to the development process. These requirements
included which propellants to use, thrust level and
engine throttling capabilities, and specific impulse and engine
life requirements. 90% H2O2 and RP-1 kerosene were
specified because of their common use in modern rocket engines,
and for use with the catalyst beds. In order to fully
leverage the turbopump design and conserve tank weight, a
maximum tank ullage pressure of 50 psi was specified
for both propellants. Sea-level thrust was specified at 5000
lbf, with a mission-averaged Isp of 260 seconds and an
engine life of 1000 seconds for adequate reusability.
Additionally, in the interest of surpassing the performance of
H
-
3
American Institute of Aeronautics and Astronautics
Figure 2. Flowchart of the design process for overall pump
assembly
Figure 1. Required pump power curve at given shaft speed
0
100
200
300
400
500
600
700
800
900
30000 40000 50000 60000 70000 80000 90000 100000
Po
we
r R
eq
uir
ed
[H
p]
Shaft Speed [RPM]
Ox Pumpn n = 1
Ox Pump n = 2
Fuel Pump n = 1
Fuel Pump n = 2
past designs, the thrust-to-weight ratio of the dry engine was
required to be at least 50% greater than the BS-605
ratio of approximately eleven. As will be shown with multiple
computational analysis methods, the developed
turbopump satisfies the overall objectives as well as all
specific requirements.
IV. Design and Analysis Methodology
Several different analysis methods were used to develop the
design. From a fluid perspective, 1-D, 2-D, and 3-D
equations and computational fluid dynamics (CFD) analyses are
used to evaluate the shape and operation of the
turbopump. These are used to set the operating point and predict
how the pump will behave in off-design operations
such as throttling or start-up. Finite element analysis (FEA)
software is used to predict failure modes as well as to
provide feedback to the overall 3-D and 2-D design and is used
to predict pump temperatures, vibrations, and
structural displacements. Once these design methodologies
converge on a design that meets all necessary operation
criteria, detailed component testing and experimental
verification then leads to final manufacturing.
Figure 2 outlines the various different design approaches. 1-D
Microsoft Excel-based codes were used for the
initial sizing and optimization of the pumps, which use 1-D
steady-state equations and empirical data to approximate
sizing and overall performance. 2-D and 3-D design was performed
in CFturbo turbomachinery design software, a
program that uses
empirical data to
create a 3-D
computer-aided design
(CAD) model of a
pumps inducer, impeller,
stator/diffuser, and
volute based on
several different
design approaches. It
is convenient because
it can be tuned to
create geometry based
on its own internal
design software, or
simply generate a
pump based on users input. The pump in this paper was designed
with a mixture of CFturbo design input as well as the 1-D codes
that
were written in Excel. XLRotor, a proprietary Excel code, was
used to generate Campbell diagrams that show modal
response of all rotating
machinery. It used inputs from
the 3-D geometry created by
CFturbo and integrated in
Dassault Systmes SolidWorks.
The overall assembly CAD
model was generated in
SolidWorks. Three-dimensional
CFD and FEA simulations were
conducted using ANSYS CFX,
ANSYS Mechanical, and NX
Nastran, and were the last steps
in the design process to validate
1-D and 2-D design codes. These
analyses also provided useful
feedback as to whether or not the
design at the beginning of the
process met all desired goals and
yielded no critical failure modes
within the design envelope.
-
4
American Institute of Aeronautics and Astronautics
Figure 3. Detailed layout of final turbopump configuration,
with
main components labeled
The iterative nature of the process evolved over the course of
the design, and each component of the pump
changed over a different number of iterations. The limited time
available in one semester did not allow for the
design to converge to a completely final solution, but the
design was instrumental as the basis of a second-
generation design that is currently being manufactured by a
group of Purdue students as an extension of the original
class. Extensive experimental validation of all predictions is
required to validate the following analyses and help
determine if the assumptions and equations used are accurate
with such a small pump.
As can be seen in
Figure 1, power requirements for both pumps drop significantly
as RPM increases, but there are diminishing
returns beyond approximately 75,000 RPM. At the beginning of the
design cycle, it was uncertain what level of
pressure rise was attainable from very physically small inducer
designs. Through preliminary research, it was
concluded it would be possible to design an inducer that could
provide at least 500 psi pressure rise through the
oxidizer pump, which would make it possible to run the turbopump
assembly at 90,000 RPM. For any smaller
pressure rise, the pump speed would have to be lowered. Despite
diminishing returns in power requirements above
75,000 RPM, an increase to 90,000 RPM allowed for a smaller,
more efficient radial turbine design that could power
the pumps. Staging would entail splitting the required pressure
rise onto two different impellers, but it is not
uncommon to see pumps that have the pressure ratio per stage
that has been designed into this system4. In order to
reduce complexity in the overall system and keep shaft length as
short as possible to create a stiffer, easier to control
rotor dynamic response, it was decided the benefits from staging
do not outweigh the increased complexity it would
entail to incorporate this detail. Also, all requirements for
engine performance were met with single-stage pumps,
and the turbine was designed to provide the power necessary
within an adequate pressure ratio margin.
V. Turbopump Design
A. Assembly Design The turbopump was designed with a single
shaft and connects two
single-stage centrifugal pumps, one for H2O2 and the other for
RP-1, to
a radial turbine driven by decomposed H2O2. The fuel pump has
a
vaned elbow axial inlet, and the oxidizer pump has a vaneless
radial
inlet. Because the fuel inlet is a 90o elbow, a vane is required
to
minimize secondary flow losses by helping maintain axial flow.
Both
pumps are fed by inducers and discharge through single-exit
volutes.
The turbine inlet consists of a dual-inlet volute and a single
exit channel
that flows directly into the chamber injector. The assembly is
mounted
to the injector inlet flange with the shaft in line with the
longitudinal
axis of the chamber.
B. Inducer Requirements and Design In the preliminary design of
both
turbopump impellers, it was determined
that a shaft speed of 90,000 RPM would
be necessary to attain acceptable pump
efficiencies. This speed is quite high,
and represents a key design driver. Of
principal concern is the issue of
excessive cavitation in pump
components at such high tip speeds,
which would result in a loss of pump
head and very poor performance. The
onset of cavitation occurs when the
available net positive suction head
(NPSH), defined in Eq. 1, falls below
the required net positive suction head
Figure 4. Final pump design with
colored sections for clarity
-
5
American Institute of Aeronautics and Astronautics
Table 2: Inducer Performance Goals
Parameter Goal Explanation
Suction
specific speed Maximized Minimize NPSHR
Flow
coefficient 0.10
(Flow velocity/tip speed)
Maximize to lessen secondary flow losses
Head
generated
6-12% of total
pump head
Generate just enough pressure rise to avoid
cavitation in impeller
Deviation < 15 Minimize impeller off-design operation
Table 1: Inducer Requirements
Parameter Requirement Component
Affected
Impeller
cavitation loss
Less than 3%
total pump head Both inducers
Inlet pressure 50 psig Both inducers
Shaft speed 90,000 RPM Both inducers
Shaft diameter 0.5 in
0.7 in
Fuel inducer
Oxidizer inducer
Inlet diameter Greater than shaft
diameter Oxidizer inducer
Outlet hub
diameter
Equal to impeller
inlet hub diameter Both inducers
Tip diameter Equal to impeller
inlet tip diameter Both inducers
Tip clearance 0.01 in Both inducers
(NPSHR), defined empirically in Eq 2a. Thus, NPSH should be
maximized and NPSHR minimized for an optimal
design. Equation 1 shows that available NPSH increases for
increased inlet pressure and axial flow speed, and
decreases for increased fluid density and exit pressure. Eq. 2a
shows that required NPSH decreases for lower tip and
axial flow speeds, as well as work done on the fluid.
(
)
(1)
(
) (
) (2a)
(2b)
(
) (2c)
In order to prevent cavitation in the impellers,
inducers were designed to increase the available
NPSH and decrease tank pressure requirements.
Several additional requirements, summarized in
Table 1, also drove the inducer design. The
overall driving requirement is a maximum total
head loss due to cavitation of no more than 3% in
the impellers, such that pump performance goals
can be met. A maximum tank pressure of 50 psig
was dictated so the pumps can run off of tank
head for a nominal tank size, thus allowing for
thinner tank walls and lower weight. The shaft
speed, as mentioned, was set at 90,000 RPM. The
shaft diameter was dictated by a vibrational stress
analysis performed in XLRotor. Both inducers
were designed to increase the available NPSH in
impellers designed a priori, thus the outlet
dimensions of the inducers were required to
match the inlet dimensions of the impellers.
Additionally, because the pumps are arranged in a cold/cold/hot
arrangement of fuel pump/oxidizer pump/turbine,
respectively, the oxidizer inducer must fit around the shaft for
machining and assembly purposes. This is depicted in
the schematic given by Figure 3. Finally, due to machining
tolerances of 0.005 in, it was determined that the
minimum realistic tip clearance possible is 0.01 in, to ensure
pump blades do not strike the casing walls. This is
larger than typical values found in turbopumps, and is limited
by machining equipment and budget restrictions.
In addition to these requirements, there were a number of
parameters that, though desired, were not mandatory.
These desired parameters, summarized in Table 2, mainly reflect
performance goals in accordance with best
practices in industry. They were used as benchmarks against
which to compare our design as opposed to hard targets
which must be satisfied. Primarily, the inducers have been
designed for maximum suction specific speed (NSS),
given by Eq. 3. By maximizing NSS, the required NPSH of the
impeller is minimized.
(3)
-
6
American Institute of Aeronautics and Astronautics
Flow coefficient, defined as the ratio of axial flow velocity to
tangential tip velocity, was set at a typical value of
0.10. This requirement helps mitigate secondary flow losses
through the tip clearance, which is relatively large due
to the aforementioned machining limitations and because the pump
components are so small. The goal of generating
between 6-12% of the total pump head is a general rule of thumb,
intended to ensure that the more-efficient impeller
is responsible for the majority of the total head generated.
Finally, the deviation, or slip, is set to less than 15o in
order to minimize efficiency losses in the impeller due to
off-design operation. This maximum requirement was set
at a relatively high value, as detailed CFD analysis is required
to accurately predict deviation.
During the preliminary design phase, a Microsoft Excel code was
developed for use in quickly obtaining design
parameters from relevant flow conditions and component geometry.
Key parameters were calculated to make use of
an S-D plot, shown in Figure 5, from NASA SP-8052, which was
used as a benchmarking map to determine whether the preliminary
design is satisfactory.
3 The map takes into account flow coefficient, a
dimensionless
cavitation number (given in Eq. 4), the suction speed corrected
for flow blockage, and the corrected suction specific
diameter of the inducer. As shown in Figure 5, both the fuel and
oxidizer inducers fall near the optimal corrected
suction specific diameter, showing that the preliminary design
is acceptable for further refinement.
(4)
Figure 5: S-D diagram for preliminary inducer design, adapted
from NASA SP-8052.3 The red diamond represents the fuel inducer,
and the blue circle represents the oxidizer inducer. Note how
closely the
preliminary design falls along the optimum suction specific
diameter curve.
C. Preliminary Inducer Performance With this preliminary design
as a benchmark, CFturbo was used to generate inducer geometry for
CFD and
FEA simulations. The CFturbo process was adapted from advice
given by Dr. Edward Bennett of Mechanical
Solutions, Inc. While powerful for initial design, CFturbo
exports geometry that requires some manipulation in
order to work properly in analysis software, particularly ANSYS
CFD and FEA software. For this reason,
SolidWorks CAD software was used to manipulate the CFturbo
output files. The final inducer designs are
summarized in Table 4.
-
7
American Institute of Aeronautics and Astronautics
D. Inducer CFD Analysis Inducer fluid domain geometry was
imported from SolidWorks, and represented the inverse of the actual
inducer
geometry: fluid areas became filled solids, and actual solid
areas were left void. A tetrahedral mesh was created
using these geometries with ANSYS Mesh. Typical meshes contained
around 1,000,000 elements. Higher element
densities were placed near blades. Fuel and oxidizer inducer
meshes are shown in Figure 6. For CAD visualization,
see Figure 7.
The meshes were imported into ANSYS CFX 14 as rotating domains.
The stationary-frame stagnation
pressure was prescribed at the inlet. The fuel inducer has a
circular inlet and the oxidizer inducer has an annular
inlet, centered around the shaft. The mass flow rate was
prescribed at the outlet. The hub and blades were set as no-
slip walls in the rotating domain. The outer wall surrounding
fluid domains are stationary; this is defined as a
counter-rotating wall in CFX. The Rayleigh-Plesset cavitation
model was selected and simulations were performed with and without
cavitation to predict how severely cavitation would decrease
inducer performance.
Liquids were assumed incompressible, and both liquid and vapor
phases were assumed isothermal at 25C. The
Shear Stress Transport (SST) turbulence model was used because
it has been reported to more accurately predict
flow separation than the k- model, the default option in
CFX.5
To aid with convergence, a simulation was
performed with the inducers inlet pressure set extremely high,
such that no cavitation occurred
anywhere in the domain. Once this solution
converged, the inlet pressure was gradually lowered,
and the simulation re-run with initial values specified
by the previous results. This procedure was repeated
until the inlet pressure was set to the actual desired
value. Table 3 describes the tetrahedral meshes used
for inducer simulations. Simulation inputs and outputs
for the inducer analysis are shown in Table 5 and
Table 6, respectively. Because an additional design
iteration occurred after these analyses were
performed, and time did not permit additional
analysis, some parameters do not match those of the
final design.
Table 4: Inducer design summary
Parameter Unit Oxidizer
Pump
Fuel
Pump
Diameter in 1.07 0.74
Length in 0.50 0.73
Rotation speed RPM 90,000 90,000
Mass flow rate lbm/s 17.64 3.31
Specific speed (US) 5408 2678
NPSHR estimate ft 560 253
Flow coefficient - 0.11 0.089
Head coefficient - 0.074 0.073
P generated
psi
% total
580
10.8%
269
6.2%
Cavitation P loss psi 50 3
Axial Thrust lbf 325 127
Table 3: Tetrahedral mesh cell count for
inducer simulations
Number of elements
Fuel inducer Oxidizer inducer
1,040,125 804,341
a. Fuel inducer b. Oxidizer Inducer Figure 6: Inducer CFD
meshes
-
8
American Institute of Aeronautics and Astronautics
Figure 7 displays the static pressure contours over the blades
and hub of both inducers, obtained by the
simulation in ANSYS CFX. As shown, the oxidizer inducer raises
the H2O2 to a higher pressure than the fuel
inducer raises RP-1. Figure 8 shows streamlines through the
inducers. These streamlines, and associated velocities,
are relative to the rotating reference frames of the
inducers.
a. Fuel inducer b. Oxidizer inducer
Figure 7: Static pressure contours on the inducer hubs and
blades
Table 6: Inducer CFD simulation inputs
Parameter Fuel Inducer Oxidizer Inducer
Rotational
speed 90,000 rpm 90,000 rpm
Fluid RP-1 90% H2O2
10% water
Inlet stagnation
pressure
50 psi (stationary
reference frame)
60 psi (stationary
reference frame)
Inlet swirl angle 0 degrees 0 degrees
Outlet mass
flow rate 13.2 lbm/s 3.3 lbm/s
Table 5. Inducer CFD simulation outputs
Parameter Fuel
Inducer
Oxidizer
Inducer
Outlet stagnation
pressure 319 psi 640 psi
Stagnation pressure rise
(considering cavitation) 269 psi 580 psi
Pressure rise lost due to
cavitation 3 psi 50 psi
-
9
American Institute of Aeronautics and Astronautics
Figure 9 shows a visualization of cavitation near the blade
surfaces. Red areas indicate non-cavitating regions;
all other colors indicate varying degrees of cavitation. The
fuel inducer, which produces almost no cavitation,
benefitted from a relatively open design envelope in which the
hub contour could be gradually changed for a smooth
pressure increase. The cavitation that does occur in the fuel
inducer is most severe on the suction side of the blades,
which is likely due to the relatively high angle attack of the
fuel inducer blades. The less-robust oxidizer inducer,
constrained by its position in the middle of the turbopump,
produces moderate cavitation. In both inducers,
cavitation begins at leading edges and decreases in the
stream-wise direction along the blades.
E. Pump Impeller Design and Detailed Design Methodology A
shrouded impeller design was considered at the onset of the design
process in order to cut down significantly
on tip leakage flows, as well as to strengthen the individual
blades. As a consequence of choosing a shrouded
design, there is viscous fluid shear from the interaction
between the shroud and the casing wall, which was taken
into account in a CFturbo power requirement analysis. Additional
analysis is required to determine if an
unshrouded design would reduce power requirements. From review
of the rocket engine turbopump literature,
a. Fuel inducer b. Oxidizer inducer
Figure 8: Flow streamlines through the inducers
a. Fuel inducer b. Oxidizer inducer
Figure 9: Cavitation visualization near inducer blades
-
10
American Institute of Aeronautics and Astronautics
pumps of similar size and power output tend to use a shrouded
design almost exclusively, most likely because of the
extremely large blade stresses resulting from a lightweight
design imparting an incredible amount of momentum to
the working fluid.
For off-design performance standpoint, three different
computational methods were used to predict how the
pumps would behave at off-nominal rotational speeds,
backpressures, and flow rates. CFturbo was used to account
for the losses of the pump impeller, diffuser, and volute, and
uses affinity laws to approximate the pump
performance. There is no prediction of stall, surge, or
cavitation losses, but it does take into account frictional
losses
in the pump, as well as scaling of volute losses from different
mass flows. It is capable of producing a pump map,
but the map seems to be quite conservative in its prediction of
the pump performance. For example, the pressure
ratio predicted does not always lie on the same throttle line or
speed line that was designed, but is sufficiently close.
CFD results were also used to generate data points at off design
operating ranges, and were run for a number of
cases that are described in more detail in the following
section.
The third method of analyzing off-design performance was by
using the Two-Elements in Series (TEIS) method outlined by
Japikse.
4 Japkises model uses a combination of about 50 equations
solving for approximately
50 variables (depending on what assumptions are made in the
analysis), some of which are energy and mass flow
equations, whereas several are empirical equations that are used
to create relationships between variables that allows
for the system of equations to be solved.
A total of four fluids are solved in a control volume approach,
utilizing a commonly accepted jet/wake impeller
exit model, where the fluid exiting the impeller consists of a
primary (jet) fluid, denoted by subscript p, with a
higher velocity than the secondary (wake) fluid, denoted by
subscript s. The other two fluids used in equations to
solve for the impeller flow are the inlet flow,
and a hypothetical flow that consists of the jet
and wake flow after they have been fully mixed,
denoted by subscript m. This approach is shown
schematically in Figure , from Japikse.
In solving for the flow states of all four
fluids, one can know the entire flow
characteristics of the impeller, as well as
directly calculate efficiency. The analytical
model is a 2-D CFD approach where flow
conditions at both the inlet and the exit of the
impeller are solved for continuity of mass flow
and rothalpy, a term defined in Eqs. 5 and 6.
Equation 5 is a term that defines outlet rothalpy,
and Eq. 6 defines inlet rothalpy for the impeller.
Rothalpy is a contraction of rotational
stagnation enthalpy, and is a term that takes into
account the change in enthalpy of the fluid due
to the rotation of the impeller wheel.
(5)
(6)
By varying , the relative velocity component of the primary mass
flow stream, and ,the mass flow ratio between the primary and
secondary flow, one is able to conserve mass flow through all four
fluid components. This
is a highly iterative approach, and is dependent on a code that
is written to make sure that all energy and mass flow
conservation equations are satisfied, and that all resultant
values line up as close as possible to chosen experimental
values.
Figure 10. Diagram of jet/wake model as viewed along the
exit plane of an impeller wheel.4
-
11
American Institute of Aeronautics and Astronautics
The TEIS model can be used to either optimize geometry or simply
solve for off-design conditions by varying
the rotational speed (rothalpy will change), or the mass flow
input, and it will be able to calculate resultant
efficiencies and pressures. This approach was not used in the
overall system off-design analysis because the code
was finished too late in the design
process, but it has promise to assist in
future design iterations. When compared
to CFD data, it seems to agree well. When
normalizing a pump map by optimal
design pressure and mass flow rate, one
can draw the conclusion that the TEIS
model and CFD results agree quite well,
whereas the CFturbo model does not.
However, it must be taken into
consideration that CFturbo calculates
losses in the volute, which was not
considered in the CFD or TEIS model, but
could be if more time was permitted to
run these models. Each point of the TEIS
model takes about 20 minutes to calculate
using an excel-based genetic SOLVER
algorithm that was built into the program.
The results are shown in Figure 10. The
final impeller 1-D pump parameters are
summarized in Table 7.
F. Pump Impeller CFD Analysis CFD simulations were performed for
the impellers in the same manner as the inducers. CFD results
closely
matched TEIS results, but CFturbo predicts significantly lower
pump outlet pressures, most likely because its
empirical relations only apply for rotational speeds lower than
90,000 RPM. The performance targets set for the
project can be met according to the fluid analyses, as seen in
Figure 11. However, FEA results indicate that both
pump and turbine blade designs could benefit from further
optimization.
Table 7. Final impeller 1-D pump parameters
Parameter Fuel Oxidizer Units
Mass Flow Rate 3.3 17.6 lbm/s
Pressure Rise 4496 5946 psi
Rotational Speed 90000 90000 RPM
Stage Specific Speed 455.7 911 (US)
Overall Pump
Efficiency 0.29 0.513 -
Power Required 241 0.504 HP
Chosen Pump Head
Coefficient 0.65 588 -
Shaft Diameter Chosen 0.50 0.75 in
Exit Diameter 2.36 2.56 in
Figure 10. TEIS model impeller map results
-
12
American Institute of Aeronautics and Astronautics
Figure 11. Pressure and velocity contours for oxidizer pump
impeller and volute
G. Turbine Design The turbine design was developed via similar
methods as the impellers and inducers, through 1-D calculations
as
well as 2-D and 3-D CAD and proprietary design software. For
brevity, the full turbine design process will not be
addressed in this paper. However, the analysis of the turbine
performance is useful in order to understand and
appreciate how the turbopump meets design goals.
The main drivers of the impellers are mass flow and output
pressure requirements. The density of the fluids is of
secondary importance to the hydrodynamic design, as they are
neither extremely dense nor thin. As a result, most of
the designing and sizing the impellers was purely with respect
to the fluid/hydrodynamic design. After the general
shape of the turbine was established, structural elements and
constraints were addressed.
H. Turbine CFD Analysis A full discussion of the turbine CFD
analysis is outside of the scope of this paper. However, a brief
overview
will serve as a good background to the work performed. The rotor
and stator were modeled in CFturbo, exported
as Turbogrid curve files, and imported into ANSYS
Turbogrid for meshing. The Automatic Topology and
Meshing (ATM) topology setting was selected for rotor
and stator meshes. The turbine volute was modeled in
SolidWorks and exported as an STL file, which was
imported to ANSYS ICEM-CFD for meshing. The grid
type and element count of each mesh is listed in Table 8.
Meshes were imported into ANSYS CFX to be
analyzed together.
The three meshes were connected in CFX using a
frozen rotor interface. A mass flow boundary condition was
specified at volute inlets, and a static pressure boundary
condition was specified at the rotor outlet. A no-slip wall
boundary condition was imposed for all wall and blade
Table 8. ICEM-CFD grid details for the turbine
simulation model
Part Meshing
Tool
Grid Type Number of
Elements
Rotor Turbogrid Hexahedral 362,336
Stator Turbogrid Hexahedral 137,819
Volute ICEM CFD Tetrahedral 178,899
Table 9. Simulation inputs for the turbine CFD simulation
Analysis Type Steady State Analysis Type Steady State
Domain
Rotor (rotating mesh)
90,000 rpm
13 blades
Stator (stationary mesh)
11 blades
Volute (stationary mesh)
Boundary
Conditions
Inlet
Location: Volute Inlet
Mass Flow Rate: 7.38 kg/s (3.69 kg/s at
each inlet)
Total Temperature: 1,033K
Outlet
Location: Rotor Outlet
Static Pressure: 26.02MPa
Physics
Definition
Fluid
H2O2
Turbulence Model: k- Heat Transfer: Total Energy
Rotor Stator
Interface Frozen Rotor
-
13
American Institute of Aeronautics and Astronautics
surfaces. Simulation inputs are outlined in Table 9.
Simulations
converged to an RMS residual of 10-4
. Performance outputs from
the simulation are presented in Table 10.
VI. Hardware Design
All components of the assembly are bolted to each other, and
then v-clamped to plumbing. In addition to main propellant
and
exhaust plumbing, several drains and high-pressure
hydrostatic
seal lines are incorporated into each component, as shown in
the
assembly cross-section in Figure 3 and Figure 12. In the
interest of
cost, ease of manufacture, and compatibility with both
propellants,
the pump inducers, impellers, and cases are made of 300
series
stainless steel. The turbine impeller is made of Nimonic 90
because of high temperatures and loads, but the shaft and
turbine case are made of Inconel 718 for lower cost. The
materials and manufacturing processes for each component are
detailed in the cross section view provided by Figure
12.
Figure 12. Detailed layout of the final turbopump configuration,
with component materials and
manufacturing processes labeled
Table 10. Performance outputs from the
turbine CFD simulation
Total Pressure Ratio 1.64
Total Temperature Ratio 1.04
Torque 64.6
N-m
Power 609
kW
Total Isentropic
Efficiency 72.5%
-
14
American Institute of Aeronautics and Astronautics
To accommodate complex design features and reduce manufacturing
cost and time, additive manufacturing via
direct metal laser sintering (DMLS) was identified as the best
process. The relatively small size of the turbopump
allows the use of DMLS, which is only currently possible in
printers with a build envelope considerably smaller than
many conventional turbomachinery components. DMLS is a rapidly
developing process, and material specifications
can be brought close to that of cast parts, in terms of both
density and strength. A scale model was constructed out of
a polymer in a fused deposition modeling printer, shown in
Figure 13, to give a sense of scale and to serve as an
educational display.
VII. Conclusion
In order to verify the performance of the turbopump, it is
desired to construct a test rig that is able to verify all
computational models and ensure that the turbopump meets all
requirements. One way to test the predictions of
pump performance would be to conduct incremental tests where a
turbopump is constructed and first tested in a cold
flow environment, where the outlets of the pumps are vented to
ambient conditions. Pressure would be measured at
the exit of the impeller to verify TEIS and CFD model
predictions of performance, and allow for the tailoring of
each predictive method. Also, cavitation in the inducer section
of the pump could be verified by an axial line of
pressure sensors. If a major pressure drop is detected by a
sensor, it may be an indication of cavitation. By running
the pump impellers or inducers for a significant amount of time,
one could also infer the resistance of the impellers
to cavitation by post-flow examination.
Such a test rig has been designed and is currently being
manufactured for testing at the Maurice J. Zucrow Labs.
The pump will be cold flow tested with water by a single
impeller and inducer, driven by flowing heated air over a
turbine. The test rig will be initially instrumented to verify
flow rates and inlet and outlet pressures, and will include
visual access to the inducer to investigate cavitation via high
speed video. After a significant effort in testing the
turbopump and validating the design, the turbopump will be
further developed to either pump H2O2 for a rocket
engine application, or inform a new design for a dual-impeller
bipropellant turbopump. This work will thus give way
to a first capability of academic turbopump research and design
at Purdue University.
VIII. Acknowledgments
Special thanks go to Prof. Steve Heister for establishing this
course and for his patience and enthusiasm
throughout the design process, Dr. Edward Bennett and Travis
Jonas of Mechanical Solutions, Inc. for their
expertise and substantial guidance in turbomachinery design,
John Munson of Rolls-Royce for his expert advice on
high-speed bearings, and finally Dr. Robert Frederick, Gene
Fleeman, and the ASEE Propulsion Education
Committee for investing their time in our endeavors.
Figure 13. A picture of the full-scale 3D prototype before
painting
-
15
American Institute of Aeronautics and Astronautics
References
1. Blank, R. A., Pourpoint, T. L., Meyer, S. E., Heister, S. D.,
Anderson, W. E., Experimental and Theoretical Performance of High
Pressure Hydrogen Peroxide Catalyst Beds, Journal of Propulsion and
Power, Vol. 28, pp. 912-917, 2012.
2. Hill, Philip G., and Carl R. Peterson. Mechanics and
Thermodynamics of Propulsion. Reading, MA:Addison-Wesley, 1992.
Print.
3. Jacobsen, J. K., and Keller, R. B., Jr., Liquid Rocket Engine
Turbopump Inducers, NASA SP-8052, 1971. 4. Japikse, David, William
D. Marscher, and Raymond B. Furst. Centrifugal Pump Design and
Performance.
Wilder, VT: Concepts ETI, 1997. Print.
5. Menter, F. R., Improved two-equation k-omega turbulence
models for aerodynamic flows, NASA STI/Recon Technical Report N,
vol. 93, p. 22809, Oct. 1992.